Back to skill

Security audit

时光梳理·照片视频梳理

Security checks for vulnerabilities and agentic risk

Overview

This photo organizer has file-handling flaws that can add unapproved move decisions or move files from outside your selected library.

Work on backed-up copies and avoid bulk moves until CSV quoting and library-boundary checks are fixed. Cloud uploads preserve photos and their location/time metadata, so choose the destination deliberately. Keep any automatic task monitor limited to the current job.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
.
Finding

Unquoted review CSV enables injected movement decisions and directory traversal

Content
View full analysis

Impact

With root /work/library and subdirectory quarantine, the injected ../private.txt resolves to source /work/private.txt and destination /work/library/private.txt. If the source exists, the destination does not, and filesystem permissions permit it, the script relocates the unrelated file into the library. This can disrupt files outside the selected library and bring unrelated content into later library export or upload operations.

Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (8)

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read, file_write, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The cloud-import workflow lacks a warning about transmitting photos and their location metadata to third-party services.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

Phase 8 instructs importing the processed library into phone or cloud albums after earlier phases add GPS and repair capture timestamps. The skill provides backup and rollback safeguards but does not disclose that cloud import transmits potentially sensitive media and enriched location history to a third-party provider.

Content

No source excerpt is available for this finding.

The cloud-upload workflow lacks a privacy warning about transmitting photos, videos, and preserved location metadata to a third-party provider.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

This workflow instructs uploading original media to cloud storage while preserving metadata, after earlier stages explicitly add GPS coordinates inferred from travel records. It warns about metadata loss but does not warn that the cloud provider will receive sensitive media, capture times, and location history.

Content

No source excerpt is available for this finding.

The example workflow directs cloud import of photos without warning about third-party disclosure of photos and embedded location metadata.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

This is an actionable workflow step directing import through Baidu Netdisk, not merely a catalog of available services. The preceding workflow adds itinerary-derived GPS to EXIF, but the file provides no privacy warning that cloud upload can disclose both the images and their location metadata to the provider; hash verification only addresses file integrity.

Content

No source excerpt is available for this finding.

The import workflow instructs uploading the photo library to Google Photos without a privacy warning.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

This walkthrough explicitly selects a Google Photos upload, rather than merely listing an available import target. Uploading originals transmits personal photos and retained metadata, potentially including itinerary-derived GPS locations, to a third-party service; the file discusses metadata preservation but does not disclose this privacy consequence.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/exif_scan.py (reported line 36)May include surrounding context.

python
for i in range(0, len(todo), B):
        batch = todo[i:i+B]
        # -CreateDate 兼容视频/照片; -DateTimeOriginal 照片; -GPSLatitude/-GPSLongitude
        res = subprocess.run([EXIFT,'-T','-n','-CreateDate','-DateTimeOriginal',
                              '-GPSLatitude','-GPSLongitude','-FileSize'] + batch,
                             capture_output=True, text=True)
        for p, line in zip(batch, (res.stdout or '').strip().split('\n')):

Tainted flow: 'batch' from open (line 34, file read) → subprocess.run (code execution)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/exif_scan.py (reported line 36)May include surrounding context.

python
for i in range(0, len(todo), B):
        batch = todo[i:i+B]
        # -CreateDate 兼容视频/照片; -DateTimeOriginal 照片; -GPSLatitude/-GPSLongitude
        res = subprocess.run([EXIFT,'-T','-n','-CreateDate','-DateTimeOriginal',
                              '-GPSLatitude','-GPSLongitude','-FileSize'] + batch,
                             capture_output=True, text=True)
        for p, line in zip(batch, (res.stdout or '').strip().split('\n')):

The trigger “统一文件名” (“standardize filenames”) is not scoped to photos or videos.

Low
Category
Trigger Abuse
Confidence
88% confidence
Finding

This trigger also matches ordinary requests to rename documents, source files, or other non-media files. Although the description identifies a photo/video organizer, the trigger itself has no media qualifier or exclusion condition to prevent unrelated file-renaming requests from activating it.

Content

No source excerpt is available for this finding.